-
Notifications
You must be signed in to change notification settings - Fork 1
/
spectral_analysis.py
180 lines (165 loc) · 5.49 KB
/
spectral_analysis.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
#!/usr/bin/env python
# ~*~ coding: utf8 ~*~
"""Spectral analysis of ameriflux - CASA differences."""
import matplotlib as mpl
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
from statsmodels.multivariate.pca import PCA
# This is probably what my screen is
mpl.rcParams["figure.dpi"] = 144
mpl.rcParams["savefig.dpi"] = 300
corr_data1 = pd.read_csv(
"ameriflux-minus-casa-half-hour-towers-autocorrelation-functions.csv", index_col=0
)
corr_data2 = pd.read_csv(
"ameriflux-minus-casa-hour-towers-autocorrelation-functions.csv", index_col=0
)
corr_data = pd.concat([corr_data1, corr_data2], axis=1)
# corr_data = corr_data2
corr_data.index = pd.TimedeltaIndex(corr_data.index)
corr_data.index.name = "Time separation"
corr_data = corr_data.astype(np.float32)
pair_counts1 = pd.read_csv(
"ameriflux-minus-casa-half-hour-towers-pair-counts.csv", index_col=0
)
pair_counts2 = pd.read_csv(
"ameriflux-minus-casa-hour-towers-pair-counts.csv", index_col=0
)
pair_counts = pd.concat([pair_counts1, pair_counts2], axis=1)
# pair_counts = pair_counts2
pair_counts.index = pd.TimedeltaIndex(pair_counts.index)
HOURS_PER_DAY = 24
DAYS_PER_YEAR = 365.2425
HOURS_PER_YEAR = HOURS_PER_DAY * DAYS_PER_YEAR
N_YEARS_DATA = 5
TOWERS_LONG_DATA = [
name
for name in corr_data
if corr_data[name].count() >= HOURS_PER_YEAR * N_YEARS_DATA
and pair_counts.loc[
slice(None, "{0:d} days".format(int(DAYS_PER_YEAR * N_YEARS_DATA + 1))), name
].min()
>= HOURS_PER_YEAR * 0.5
]
uniform_corr_data = corr_data.loc[
slice(None, "{0:d} days".format(int(DAYS_PER_YEAR * N_YEARS_DATA))),
TOWERS_LONG_DATA,
]
uniform_corr_data.plot(
subplots=True,
sharey=True,
figsize=(12.5, 6.5),
title="AmeriFlux - CASA autocorrelations",
xticks=pd.timedelta_range(start=0, freq="365D", periods=N_YEARS_DATA + 1)
.to_numpy()
.astype(float),
)
plt.subplots_adjust(bottom=0.07, top=0.95, left=0.05, right=0.95)
fig = plt.gcf()
for ax in fig.axes:
ax.set_xticks(
pd.timedelta_range(start=0, freq="365D", periods=N_YEARS_DATA + 1)
.to_numpy()
.astype(float),
minor=False,
)
ax.set_xticklabels(
["{0:d} years".format(year) for year in range(N_YEARS_DATA + 1)],
rotation=0,
minor=False,
)
fig.savefig("ameriflux-minus-casa-long-correlation-data.png")
uniform_spectrum = pd.DataFrame(
np.fft.rfft(uniform_corr_data, axis=0),
columns=uniform_corr_data.columns,
index=np.fft.rfftfreq(uniform_corr_data.shape[0], 1.0 / HOURS_PER_DAY),
)
uniform_spectrum.index.name = "Freq. (1/day)"
abs(uniform_spectrum).plot(
subplots=True,
sharey=True,
logy=True,
figsize=(12.5, 6.5),
xlim=(0, 6),
title="AmeriFlux - CASA CO$_2$ Flux difference spectra (log scale)",
)
plt.subplots_adjust(bottom=0.07, top=0.95, left=0.05, right=0.95)
plt.savefig("ameriflux-minus-casa-long-correlation-spectrum-days.png")
uniform_spectrum_year = pd.DataFrame(
np.fft.rfft(uniform_corr_data, axis=0),
columns=uniform_corr_data.columns,
index=np.fft.rfftfreq(uniform_corr_data.shape[0], 1.0 / HOURS_PER_YEAR),
)
uniform_spectrum_year.index.name = "Freq. (1/year)"
abs(uniform_spectrum_year).plot(
subplots=True,
sharey=True,
logy=True,
figsize=(12.5, 6.5),
xlim=(0, 8),
title="AmeriFlux - CASA CO$_2$ Flux difference spectra (log scale)",
)
plt.subplots_adjust(bottom=0.07, top=0.95, left=0.05, right=0.95)
plt.savefig("ameriflux-minus-casa-long-correlation-spectrum-years-log.pdf")
abs(uniform_spectrum_year).plot(
subplots=True,
sharey=True,
logy=False,
figsize=(12.5, 6.5),
xlim=(0, 8),
ylim=(0, 1000),
title="AmeriFlux - CASA CO$_2$ Flux difference spectra (linear scale)",
)
plt.subplots_adjust(bottom=0.07, top=0.95, left=0.05, right=0.95)
plt.savefig("ameriflux-minus-casa-long-correlation-spectrum-years-linear.pdf")
plt.pause(1)
pca_results = PCA(
uniform_corr_data.iloc[::4, :],
# This should keep correlation components centered on zero
demean=False,
# Data is already standardized
standardize=False,
missing="drop-min",
ncomp=10
)
# Boxplots of R^2 for tower ~ comp0, tower ~ comp0 + comp1, ...
pca_results.plot_rsquare()
plt.savefig("ameriflux-minus-casa-long-correlation-pca-rsquared.pdf")
# Essentially differencing the plot above
pca_results.plot_scree()
plt.savefig("ameriflux-minus-casa-long-correlation-pca-scree.pdf")
# In both those plots, look for elbow.
pca_results.scores.plot(
subplots=True,
sharey=True,
ylim=(-0.03, 0.03),
title="Principal compontents of correlations",
)
plt.savefig("ameriflux-minus-casa-long-correlation-pca-principal-components.png")
spectral_pca_results = PCA(
uniform_spectrum.iloc[: uniform_spectrum.shape[0] // 4, :],
# This should keep correlation components centered on zero
demean=False,
# Data is already standardized
standardize=False,
missing="drop-min",
ncomp=10
)
# Boxplots of R^2 for tower ~ comp0, tower ~ comp0 + comp1, ...
spectral_pca_results.plot_rsquare()
plt.savefig("ameriflux-minus-casa-long-correlation-spectral-pca-rsquared.pdf")
# Essentially differencing the plot above
spectral_pca_results.plot_scree()
plt.savefig("ameriflux-minus-casa-long-correlation-spectral-pca-scree.pdf")
# In both those plots, look for elbow.
spectral_pca_results.scores.plot(
subplots=True,
sharey=True,
ylim=(-0.03, 0.03),
title="Principal compontents of correlations",
)
plt.savefig(
"ameriflux-minus-casa-long-correlation-spectral-pca-principal-components.png"
)
plt.pause(1)